Sunday, November 15, 2015

Another Unsound Attack On Proposed Senate Reforms

Advance Summary

1. A recent article by a former NSW politician argues that proposed Senate reforms will create an effectively first-past-the-post system that advantages the Coalition and eliminates minor party candidates and independents.

2. The article exaggerates the impact of the proposed system on minor party candidates, since minor party candidates would have won at least three seats in 2013 under the proposed system.

3. While parties polling very low vote shares would not win without group ticket preferencing, this is not specifically because many votes would exhaust.  Rather, it is because strong preference flows between obscure parties would not exist even if all voters assigned their own preferences.

4. The article's claims about the impact of exhausting votes on preference transfers from Green to Labor and vice versa are undermined by those transfers being much less often important in Senate than in House elections.

5. The article's assumption that it would always be an advantage to run joint tickets rather than split tickets (eg for the Liberals and Nationals, or for Labor and the Greens) is incorrect.  Whether it would be better to run joint or split tickets would vary depending on party vote levels.

6. There is simply no reliable evidence that proposed reforms disadvantage any of the Coalition, Labor or the Greens, or any other force with serious support in any given state.  They disadvantage who they are designed to disadvantage: preference-harvesters.


This article deals with a recent attack on proposed Senate reforms published by New Matilda.  The article, by former NSW Legislative Councillor Peter Breen, is riddled with mistakes and contains little that is new to the debate apart from new errors.  However, some of the errors are instructive. I think Breen's article is well worth subjecting to the usual treatment for such pieces, because it is such a good example of how easy it is to make plausible-sounding and concerning, but incorrect arguments against proposed Senate reforms.  Unfortunately, Breen's article is the only public life I've seen in the Senate reform debate at all for almost two months since the last instalment.  The lack of public progress on the issue simply isn't acceptable; the 2013 Senate contest was an absolute farce, and any party which isn't serious about doing at least something to fix the mess before the next one is not worth even considering voting for.  Even if it's not the JSCEM model, it's high time that the major parties told us what reforms they'll actually support and when.

I have written quite a lot about Senate reform here; click on the Senate reform tab for all previous pieces.   As noted before, all kinds of false claims have been made about how the proposed reforms must be some kind of Coalition plot.  Actually, the reforms are deliberately designed to destroy the electoral business model of preference-harvesters, and only parties that can only win by preference-harvesting have all that much to fear.

Breen himself is a former beneficiary of the preference harvesting system - elected in the 1999 NSW LC election off 1% of the vote, the highest total of the three candidates elected at that election who would not have won under the current NSW system.  The 1999 outcome was widely seen as quite silly and led to reforms prior to the 2003 election, and in any case Breen's party/parties never much troubled the scorers again.

The proposed Senate reform model Breen's article attacks is the model recommended by the Joint Standing Committee on Electoral Matters (JSCEM).  This model has three key components.  Firstly, voters voting below the line for individual candidates will no longer need to number virtually every square. Secondly, voters will be able to distribute preferences between parties above the line.  Thirdly, automatic group ticket preferencing by parties will be abolished, meaning that if a voter just votes 1 for a party, their vote will exhaust when that party is excluded from the count.

Breen alleges that "Minor parties and independents will be decimated by the proposed changes, which will effectively abolish preferential voting and turn Senate voting into a first-past-the-post system that favours the Coalition".

There are two aspects of this argument I want to deal with in detail: firstly the alleged "decimation" of minor parties and independents and secondly the claim that the alleged first-past-the-postiness of the new system (translation: many voters choosing not to distribute preferences) would advantage the Coalition.

Minor parties and independents

Breen claims:

"The problem with eliminating the GVT is that only about 15 per cent of voters actually choose preferences – as demonstrated by the New South Wales upper house voting system – which means 85 per cent of votes for independent and minor party candidates will exhaust.

In other words, most of the 25 per cent of voters in the last Senate election who chose an independent or minor party candidate for their primary vote will find their vote in the rubbish bin before the vote count is completed. The practical effect of the proposed new voting system is to remove the independents and minor party candidates from the contest for the last seat in each of the six Australian states."

For starters, the final sentence exaggerates considerably.  Had the proposed JSCEM system existed at the 2013 Senate election and had exactly the same votes been cast for the same parties, Nick Xenophon, Xenophon's running mate Stirling Griff, Glenn Lazarus (PUP) and David Leyonhjelm (LDP) would have all been elected.  Jacqui Lambie (PUP) would have made the contest for the last seat in Tasmania, but may have lost.

The new voting system would actually have eliminated indies and micros from the final seat contests only in Victoria (Muir) and WA (Dropulich/Wang) based on the votes actually cast.  There's a strong case that in reality it would have eliminated Leyonhjelm as well, since many of the parties that contested under the current system would merge or not bother under the JSCEM proposal, reducing his ability to benefit from voter confusion about party names.  But clearly indies and "minor parties" can win under the proposed system.  Indeed, looking at the 2013 results, every party polling at least 1.2% of the vote nationwide would have won at least one Senate seat under it.  

It is of course right (both factually and democratically) that the new system would, in practice, eliminate candidates with very low primary votes from serious contention for seats.  But this is not specifically because of the low proportion of voters who might choose to distribute preferences.  Rather, it is because when micro-party voters choose to distribute preferences themselves, their preferences do not flow very strongly to any given party.  Even if 100% of micro-party voters chose to distribute their own preferences, micro-parties polling very small shares of the primary vote would still never get elected.  I demonstrated this in Optional Senate Preferencing: Not An ALP/Liberal/Green Stitch-Up.  In that article, I pointed out that when voters for micro-parties choose the distribution of their own preferences in the House of Representatives, they did not strongly favour other micro-parties generally over the "big three".

Therefore, it is not the preferencing behaviour (or lack thereof) that voters might display under the new system that would eliminate micro-parties from contention.  Rather, the cause would be the removal of party-directed group-ticket preferencing. Under the current system, a party is allowed to allocate virtually all of its preferences to another, although in fact its voters, even if all forced to make such choices themselves, would do no such thing (not even if their chosen parties asked them to!)

The exhausting of a lot of micro-party votes also won't have much impact on the "big three" (Coalition, Labor, Greens).  If voters for these parties were forced to direct their own preferences, their preferences would generally go all over the place, with little impact on the crucial contests.  If anything, it is Labor that has the most to gain from the proposed reform, since under the current system, ideological micro-parties tend to feed to the Coalition or Greens rather than to the middle.  Despite this, some of the great strategic minds of the ALP seem to think they should oppose it.

Coalition advantage from Green/ALP preferences exhausting

(Warning: this section is a bit technical, about Wonk Factor 3/5)

The other part of Breen's argument is to claim that the Coalition would benefit from having all its preferences locked within a ticket, meaning that its preferences could not exhaust.  However, when Labor and Green preferences are distributed, their preferences would tend to exhaust rather than reaching each other.  Breen argues that this advantages the Coalition.

The argument sounds logical, but it's actually incorrect, for two major reasons.  The envisaged scenario is one in which Green preferences caused a Labor candidate to beat the Coalition for the final seat (or Labor preferences caused a Green candidate to do so), but weaker preference flows between Labor and the Greens would have caused the Coalition to win instead.  This is a common scenario in the House of Representatives, but for the Senate I cannot find one single case where it has ever happened, but in which Labor and the Greens combined would not have won at least as many seats under the JSCEM proposal.  Reasons for this include that a lot of the vote total for Labor (and in some states sometimes the Greens) is locked up in the party's outright quotas (meaning that the surplus to be distributed is small), and also that Labor and the Greens are often fighting each other for a seat.

Secondly, while it's possible to construct an example in which Labor and the Greens combined would miss out on a seat they would otherwise have won because of preferences exhausting or leaking, it's also possible to construct one in which they benefit by not combining. This is very familiar territory for me as a Hare-Clark junkie, because in Hare-Clark there is a thing I call the Ginninderra Effect (warning: that link is a Wonk Factor 5.)  In said Effect, a party that "should" win based on raw party totals loses because it has only one candidate in the race, and that candidate is below the individual scores of each of two candidates from the other party.  This showed up again in Tasmania in 2014, with Labor renegade Brenton Best losing his seat because of it.   Something similar could apply under the JSCEM Senate proposal. Just as a party in Hare-Clark can be advantaged by having its remaining votes split evenly between two or more candidates, so two likeminded parties in the JSCEM Senate system could be advantaged by being on separate tickets, rather than by running a joint ticket.

Here's my artificial example, though constructing others that work the same way with quite different vote shares is easy.  In a given state the levels of support for parties are as follows (a quota is about 14.3%): Liberal 2.7 quotas, Nationals 0.7 quotas, Labor 1.6 quotas, Greens 0.6 quotas, PUP 0.4 quotas, various other micro-parties 1.0 quotas combined (with none over 0.2 quotas).

The Liberals and Nationals on the one hand, and Labor and the Greens on the other, have the choice of whether to run as a joint ticket or individually.

If both sides merge, then the Lib-Nat group (3.4 quotas) wins three, the ALP-Green group (2.2 quotas) two, and the final Lib-Nat fights PUP for the final seat.

If only the Lib-Nat side merges, then Labor wins two, Greens one, and the Lib-Nat group three.

If only the ALP-Green side merges, then the Liberals win three, the Nationals one and ALP-Green two.

If neither side merges, then barring a very strong flow of minor party preferences to Labor or the Greens, the Liberals win three, the Nationals one, Labor one and Labor and Green fight out the last.

Here's the payoff matrix in terms of seats won for each side depending on each party's decisions:

The optimum strategy for each side for the given levels of support is to run separate tickets.  The Coalition side locks in four seats by doing so.  The Labor-Green side benefits by choosing to do so only if the Coalition makes the strategic mistake of running a merged ticket.  Four seats for the Coalition with 49% of the primary vote to 31% for Labor and the Greens is quite fair given that if the "others" voters did all allocate preferences, only 30% of them would have to prefer the Coalition over ALP-Green to get the Coalition up to four quotas.  But most likely what would happen at these sorts of levels of vote support in the state is that some micros would merge and others disappear, and that as a result the Coalition's fourth seat would be threatened both by an increased Green vote and perhaps by the emergence of more broadly based alternative parties.

Why is splitting a ticket sometimes an advantage?  It is so because if one partner in a potential coalition is close to a quota then they do not need any more votes, because nobody will catch their candidate - too many votes will exhaust for that to happen.  But if that party joins with another, then votes are wasted getting one candidate on the joint ticket up to a full quota they don't actually need, and then those votes are lost to the joint ticket candidate who is fighting for the final seat.

This example does show that strategic decisions for the big parties under the proposed JSCEM system could be challenging (and there might be more challenging scenarios; for instance there might be one in which the optimum decision for each side depends on what the other side does).    But they are not nearly so challenging as the choices facing parties under the current system, and nor are they likely to involve such severe alienation of supporters as occurs when a party's best strategic choice is to make deals that contradict its own branding.

Some Other Issues

Breen's article also includes reference to "modelling by the Renewable Energy Party" (an as-yet federally unregistered party which it turns out he's "National Co-Ordinator" of) that suggests that if the JSCEM system had operated at the last three elections the numbers would be Coalition 39, Labor 25, Green 10, Xenophon 2.  No link to this modelling is given, and no evidence is presented that the REP knows what it is talking about.  Initially I could not find the modelling online at all, but I think that this (letter from Graham Askey on the REP site) might be a summary of it.  However, the "attached table" doesn't seem to be included.

In fact the conclusion is wrong, as has been shown by Antony Green here and here, and by me at the link mentioned above, in Do Proposed Senate Reforms Advantage The Coalition? and other pieces on the Senate Reform tag.  It just isn't disputable, for instance, that Glenn Lazarus would have got up under the JSCEM system.

I finally found a rather staggering explanation for the Lazarus omission deep in the comments section of the article.  The REP analysis had cherrypicked out Palmer United, claiming that they as well as the Liberal Democrats were "anomalies".  (This is understandable in the LDP case, since the LDP's election actually resulted from the flooding of the ballot paper with micro-parties.  But Breen provides no evidence at all that the election of PUP based on "a big bucket of money" is the sort of thing that is system-dependent.)  In other words the REP analysis actually at best shows what the outcomes of the last few elections would have been if those parties most inconvenient to the REP's claims about the impacts of the system had not run, but everyone else had! My bet would be that even with that taken into account, their analysis is still wrong.

Breen argues that Labor and the Greens could elect three Senators combined off a vote of 42.9%, but I have given an example above in which they might do so off only 31%!  If I change the votes in that example to 1.7 quotas for Labor and 0.7 quotas for the Greens, and reduce the Coalition vote to 46%, then ALP-Green win three seats off just 34% no matter what the Coalition does.  So Breen's bar for Labor and the Greens to potentially win three seats is set way too high.

His bar for the Coalition to potentially win four seats (at 53.4%) is also set much too high (because he ignores the possibility that running separate tickets could be an advantage).  However his claim that "The Coalition often achieves more than 53.4%", in the Senate context, is just false.  This hasn't happened in any state since 1975.  And before we get too carried away about the possibility of the Coalition winning four seats with 53.4% under the proposed system, we need to remember that in Queensland in 2004 under the current system they actually did win four with just 44.9%.  (This may well have also happened under the same vote shares in the proposed system, but the various left and centre parties would have had much better strategic resources to prevent it.)

Breen also makes the same arguments about crossbench diversity and its effects on Coalition nasties that I addressed in the previous piece by pointing out that the Senate as elected under the JSCEM model would have had much the same effects (though with a much easier negotiation process and a less "random" set of actors to deal with).

Looking further into the REP website I've found more evidence that the REP have strange ideas about how the future of parties might pan out.  For instance, in their Open Letter to Richard di Natale Part 2, Graham Askey floats a scenario in which five Senators have been elected, and the Coalition has 10.4%, the Greens 9.1% and Labor 9.08%.  But assuming that those five Senators are from these particular parties, which in REP's argument they are, then that assumes that there has been no exhaust at the exclusion of other parties at all, which can only happen if no other parties contest the election.  But this is an absurd assumption; micro-parties contest elections in the UK and USA First Past The Post systems even when they know they cannot possibly win, and new parties spring up in such systems and sometimes eventually win seats.  The number of micro-parties would reduce, but not to zero.  Most likely those that are ideologically similar to existing parties would be more likely to fold.

It seems the problem here is Breen's article doesn't spell out the extent to which Askey's analysis takes this assumption of a purely three-party system and runs with it.  The 53.4% that Breen talks about, for instance, might be 53.4% of the combined Labor, Green and Coalition vote, ignoring all others.  But ignoring all others is a silly move, since the Greens probably suffer more as a proportion of their vote from competition from likeminded micro-parties than Labor and the Coalition do.

Breen Versus Green

Finally, a common theme of Breen's article is to paint the Greens as tactically stupid and hence imply that the Greens' support for Senate reform must be a blunder too.  The history of decisions by any well established party will contain both good strategic decisions and big mistakes, so it's better to judge particular decisions on their merits.

However, the structure of Breen's arguments as concerns the Greens and Democrats is ... interesting.  Breen argues there is some kind of warning for the Greens in the fate of the Democrats, who ignored Green and independent attempts to negotiate for a directly-elected head of state at the 1998 Constitutional Convention.  Now, I'd always thought the demise of the Democrats was a product of:  an unstable leadership process, a perception of selling out over the GST, instability between the centrists and the left-of-Labor types, not being as kool with the yoof as the Greens, the fact that the Democrats had people who were in the Democrats in them, and at times just sheer bad luck.  But no, apparently the Dems succumbed because they didn't listen to the Greens and independents when the latter were proposing an idea that Breen now correctly admits was a shocker.  You learn something new every day in this game.

In reality, the Greens have correctly picked that the proposed reform is in their interests.  It might not affect their seat tally much in either direction, but it makes for a more tractable crossbench, and it stops them having to apologise to their purer supporters for making preferencing decisions that might be ideologically suspect or at times just utterly stupid.  They've also correctly picked that the crossbench's views on reform are irrelevant to the Greens.  The ball is really in the Coalition's court, since if the Coalition gets serious about reform then it will pass it with Green support, and if the Coalition isn't going to get serious then what the Greens do won't have much impact on the outcome.  About all that the Greens could do wrong here is to vote down a limited reform proposal because it doesn't go far enough - but if the government chose to introduce a very limited reform, would Labor and the crossbench really bother blocking it?  So I'm not sure what the crossbench thinks it has to talk about.

PS If anyone has cold feet about what I have written above on account of the prospect that in some rare cases Labor and the Greens would get more seats by running a joint ticket, then there is a simple alternative.  This is to adopt a variant of the JSCEM model that allows for limited group ticket preferencing - a vote that is just 1 for a party can be passed by the party to a maximum of three other parties.  This would mean large parties could make joint-ticket-vs-separate-ticket decisions more easily.  It would still allow, unfortunately, for a degree of preference-siphoning and possibly gaming attempts, but these would be a lot easier to expose.  What I am thinking is that a group of micros might hold a random draw and all allocate their second preference to the winner of that draw.  At least in this case it would be obvious who potential winners were, and easier to scrutinise them and discourage people from preferencing them unintentionally.

Added 16 Nov: See also Antony Green's article on the implications of NSW preference flows for Senate reform.  It's very detailed but the key points for the discussion above is that Antony agrees that contests between the ALP or Greens on one side and Coalition on the other will be relatively rare should the new system drive a lot of right-micro votes back to the Coalition.  Antony suggests Labor fear of the new system may be driven by concern that they can no longer use preference deals to get over the Greens for the last seat.  My comment is that with the increase in Green-preferencing micro-parties in recent elections, Labor should be more concerned about the Greens doing this to them than them being able to keep doing it to the Greens.


  1. I would have agreed with you until I saw Antony Green's analysis of how he thinks past elections could have gone under this system. That it would have basically defeated (or very plausibly defeated) every Green Senate candidate prior to 2004 (and thus would have killed the Greens in utero), raises questions about why the Greens support it now, even though their support levels are such that it benefits them now. It also would seem to drastically slash the chances of a switch such as the Greens replacing the Democrats ever happening again, short of a sudden surge that even the Greens historically couldn't pull off.

    I still kinda support in the sense that the current system is obviously flawed, and New South Wales is working, but that Green article did give me serious pause.

  2. The new system would not have defeated Bob Brown in 2001 (I suspect he would still have won then even had he not been an incumbent Senator) but would indeed likely have defeated all the other four Greens winners prior to 2004, if the same votes had been cast in an election under it. That said, one of those wins (NSW 2001) was a fluke result on One Nation preferences, as Antony notes. Cases like Tasmania 1996 and WA 1993 show that under the proposed system, if likeminded minor parties competed with each other they would risk giving seats to major parties. So if anyone is concerned about this, then they might well have reason to support a limited form of group ticketing rather than abolishing it altogether. I'm fine with that if so, because that would be a large improvement that would be less vulnerable to scare campaigns.

    However, because systems determine party strategy, I think the way these contests would have played out under the proposed new system would have been different, and that the rise of the Greens would have happened one way or other anyway. The Greens and Democrats may well have adapted to this situation in the same way the Liberals and Nationals presently minimise three-cornered Reps contests. They might agree that one would run hard in one state and dead in the other, and vice versa. It's also possible the supplanting of the Democrats by the Greens would have taken a radically different course, such as more and more "Greens" joining the party and steering it in an increasingly green direction. Because of the level of member control over policy, the Dems would have been easy to take over by force of numbers. I suspect we'd still have ended up in pretty much the same place in terms of the policies of the party most supported by left-environmentalists. We might have got there with a lot less nastiness between Greens and Democrats too.

    Some parties can draw the distinction between what helped them make it to where they are now and what might help them stay there, and some (meaning most of the current crossbench) can't. Nick Xenophon was originally elected to the SA parliament via preference harvesting, but has realised that it doesn't help him anymore.

  3. A very good article. It demolishes most of Breen`s article.

    However there is one significant error in Breen`s claim that has been overlooked. That is his claim that 85% of non-"big three" votes will not elect anyone because only 15% of voters preference.

    Firstly, he rounds down from 15.3 to 15% and then counts the votes he rounded away as not being for who they were for, rather than saying "Over 84% will exahust". Admittedly, this is a minor offence.

    Secondly, he uses only the RATL figure, despite a proportion of the BTL votes preferencing between groups. Although in his defence, more detailed research is required to find out where BTL votes go. Although, he could have used the SATL figure, which is almost all the exhausting votes just as easily.

    Thirdly, and most erroneously, he is using the figure that includes the preferences of the "big three", the largest 2 of whom on average preference at a lower rate than the overall total and represent over 73% of votes (significantly dragging down the overall figure). Most of the non-"big three" preference at a significantly higher rate than 15%. However a proportion of non"big three" RATL votes would be preferencing other non-"big three" groups.

    Fourthly, preferencing advice and behaviour would likely differ in the Senate, with its lower quota.

  4. Thankyou for adding these points. I also expect likeminded parties would be forced to find more innovative ways of arranging preference flow under the new system, and that these would increase the preferencing rate. That link of Antony's is an excellent read on how card design can affect voting rates. There's also the voter education aspect: some voters think that just voting 1 in NSW results in a full preference being distributed, since that's how it works in the Senate. If they're told it no longer works that way anywhere, some of them will learn.

  5. From a savings-provision perspective, I support the limited GVT.

    People are very much used to just voting 1. Having a limited GVT should greatly increase the chances of micro preferences having an impact on the final seat.

    The aim and implementation of any limited GVT should be that every JV1 ATL voter lists as many candidates as a BTL voter would be required to. This actually means only two other parties need be listed (five in a DD).

    Regarding micros, I expect there will be at least two rough groupings come the next election. Progressive left (centred on the Futurists) and libertarian right (centred on Leyonhjelm).